Agentic Assistant Product, Rails Platform + LangGraph Reasoning Agent
Overview
An internal agentic-assistant product combining a Ruby on Rails 7 web application with a Python LangGraph reasoning-and-action agent service. The Rails side delivers the product UX and account/data layer; the LangGraph service provides tool-using, multi-LLM agents that can call external tools and coordinate through a supervisor pattern.
Why It Exists
Delivering a production AI assistant requires both a robust application platform (auth, billing, UI, persistence) and a flexible agent runtime that can reason, call tools, and switch between LLM providers. This repository pairs the two as complementary services.
What We Built
Two coordinated components: (1) web-app, a Rails 7.2 application using Hotwire (Turbo/Stimulus), Vite, PostgreSQL, Puma, Rack::Attack, and Capistrano deployment; (2) react-agent, a LangGraph service built from the ReAct-agent template, wired to multiple LLM providers (OpenAI, Anthropic, Google Gemini, Fireworks), Tavily search, and Composio tool integrations, with langgraph-supervisor for multi-agent coordination. The agent service ships with a Makefile, Dockerfile, Kubernetes manifests, and LangGraph Cloud configuration for deployment.
Technologies & Approach
Rails 7.2 + Hotwire for a fast, server-rendered product surface; LangGraph for the agent graph with a supervisor orchestrating tool-using sub-agents; provider-agnostic LLM wiring so the assistant can route across OpenAI, Anthropic, and Gemini. Composio supplies a broad external-tool surface; Tavily handles web retrieval.
Outcome / Impact
Shows the studio can ship a complete agentic product spanning a classic application stack and a modern LLM agent runtime, with real deployment tooling (Docker, Kubernetes, LangGraph Cloud). Validates multi-LLM, tool-using, supervisor-coordinated agent architecture in production form.
Capabilities Demonstrated
- Full-stack AI product delivery (Rails 7 + Python agent service)
- LangGraph ReAct agents with supervisor multi-agent orchestration
- Provider-agnostic multi-LLM integration (OpenAI / Anthropic / Gemini / Fireworks)
- External tool integration via Composio and Tavily
- Containerized + Kubernetes / LangGraph Cloud deployment